Model reference · open weights
MiniMax-H3 is an open-weight video model from Zane91. MiniMax-H3 (BF16) weighs 498 GB; the smallest configuration that runs it is B200 180 GB.
What it is
| Released by | Zane91 |
|---|---|
| Type | Video models |
| Task | Image→video |
| Runs with | minimax-h3 |
| Released | 2026-09-26 |
| Popularity | 562 downloads / month |
| Weights | 498 GB (MiniMax-H3 (BF16), file size) |
| Licence | Its own licence terms |
What it runs on
Weights 498 GB (file size) · its biggest part 144 GB · overhead about 537 MB.
| Card | The weights | Counted memory |
|---|---|---|
| RTX 3060 12 GB … H200 141 GB | does not fit | |
| B200 180 GB | fits (encoders offloaded) | 176 GB |
Estimates, not measurements: the weights are the build's file size; a video's working memory grows with its resolution and length and is not estimated yet. diffusers can also place a pipeline's parts on separate cards (device_map) — not estimated here. Counted memory is 92 % of what CUDA reports for the card.
From the model card
Offical skills to improve prompt writing: skills on github
Use MiniMax-H3 directly via API.
Use MiniMax-H3 directly via App.
MiniMax H3 is a general-purpose, omni-modal generative system. It supports unified understanding of multimodal contexts composed of text, images, video, and audio, and can generate video with native stereo audio at resolutions up to 2K and durations of up to 15 seconds. Thanks to its task-generalization-oriented system design, H3 already possesses broad multimodal context understanding and generation capabilities at the pre-training stage, enabling outstanding performance in following complex multimodal instructions.
H3 supports the following input and output specifications:
| Category | Specification |
|---|---|
| Output duration | 4–15 seconds |
| Output aspect ratio | Supports a wide range of aspect ratios, including but not limited to 21:9, 16:9, 4:3, 1:1, 3:4, and 9:16 |
| Output resolution | Supports various resolution dimensions. The shorter side is set to 768 pixels by default. 2K | generation can be achieved with H3-Regenerate-2K |
| Output frame rate | 24 FPS |
| Output audio | 32 kHz stereo |
| Supported dialogue languages | Stable support for 11 languages: Arabic, Chinese, English, French, German, Italian, Japanese, Korean, Portuguese, Russian, and Spanish. Additional languages are also supported to varying degrees |
| Model Variant | Input Mode | Specifications |
|---|---|---|
| H3-Base-FL2VA | First-and-last-frame mode | Supports zero, one, or two input images. - No image input: Text-to-video mode - One image input: First-frame-to-video or last-frame-to-video generation - Two image inputs: First-and-last-frame-to-video generation |
| H3-Base-Ref2VA | Omni-reference mode | Supports multi-modal reference inputs: - Images: ≤ 9 images - Videos: ≤ 3 clips; each clip must be 2–15 seconds long; total duration ≤ 15 seconds - Audio: ≤ 3 clips; each clip must be 2–15 seconds long; total duration ≤ 15 seconds - Mixed inputs: Maximum number of files across all input types is 12 |
The complete H3 system consists of the following three modules:
H3-Context-IR is a hosted preprocessing and orchestration system designed for free-form multimodal inputs.
It interprets the relationships among text, images, audio, and reference videos, as well as how these materials relate to the intended generation output. Its internal workflow includes instruction parsing, cross-modal association, temporal understanding, and complex logical reasoning.
H3-Context-IR serializes its understanding of the context into a structured representation accepted by H3-Base. Without deviating from the user’s original intent, it may also supplement missing or underspecified semantic details where appropriate.
Because H3-Context-IR relies on a multi-stage workflow and multiple hosted models and services, it is not included in this open-source release. We provide an API that enables users to reproduce the behavior of the official workflow. We also provide detailed tutorials, and developers can follow the Prompting Guidance to build their own preprocessing systems.
For detailed usage instructions, see Recommended Workflow — Full 2K Workflow.
Safety Guardrails
User-submitted text, images and videos, as well as enhanced prompts, are subject to automated moderation. Content suspected of being unlawful, pornographic, or infringing third-party rights may be blocked. We use industry-standard filtering measures but cannot eliminate false positives or false negatives. These guardrails do not affect the Licensee’s obligations under the MiniMax H3 Community License, especially those relating to lawful use and use restrictions.
H3-Base encodes different modalities using their corresponding encoders or VAEs and organizes the encoded representations into a unified packed multimodal sequence. RoPE is used to capture the necessary spatial and temporal relationships among tokens before the entire sequence is passed to the H3-Omni-Transformer.
Specifically, text is encoded by the H3-Encoder; visual inputs are encoded by both the H3-Encoder and the H3-VisualVAE; and audio is encoded solely by the H3-AudioVAE.
The H3-Omni-Transformer jointly predicts video and audio latents, which are then decode
Quoted from the model card on Hugging Face — the full card is behind the Hugging Face link above.
Running it yourself
Rent a machine by the hour. Runs as it is with diffusers — on the machine, in Python.
# on your rented machine: pip install diffusers transformers accelerate ftfy
import torch
from diffusers import DiffusionPipeline
from diffusers.utils import export_to_video
pipe = DiffusionPipeline.from_pretrained("Zane91/MiniMax-H3", torch_dtype=torch.bfloat16).to("cuda")
frames = pipe(prompt="a drone shot over a forest at sunrise").frames[0]
export_to_video(frames, "/workspace/out.mp4", fps=16)